an evaluation of left and right brain dominance using … · 2020. 11. 23. · brain dominance...
Post on 20-Feb-2021
6 Views
Preview:
TRANSCRIPT
-
Abstract—The left and right brain dominance theory has been
established for decades. Besides, the left and right brain
balancing education concept and training have also been
developed for years. Currently, the only way to determine a
person whether is left or right brain dominance is by making a
questionnaire assessment. There is no scientific data that can
directly reflect brain activity to prove the left and right brain
theory as well as the effectiveness of the left and right brain
development training. Hence, in this research, it is aimed to
determine whether the electroencephalography (EEG) signal
has any correlation with the brain dominance level. The brain
dominance level of the subject is determined and benchmarked
by using the Hermann Brain Dominance Instrument (HBDI)
test, a popular testing tool utilized by innumerable multinational
companies to determine employee’s brain dominance level. As
the captured raw EEG signal is complicated and noisy, several
preprocessing methods are utilized to eliminate the unwanted
noise and artifacts efficiently from the acquired signal. The
techniques are namely baseline correction method, electrical
line noise removal, and independent component analysis (ICA).
Besides, significant features can be hardly determined from the
time-based EEG signal with high complexity. Hence, the EEG
Topographical Power Spectral Density Percentage
(EEGTPSDP) method is implemented to analyze the EEG
signal. By using the results computed by EEGTPSDP method, it
proves that there is a strong correlation between the brain
dominance level and EEG power spectral density on one
hemisphere. Hence, this research is able to validate the left and
right brain dominance theory according to the EEG signal. The
implemented EEGTPSDP method can be used to classify the
dominant brain of a person. In this way, this research is able to
contribute to the education field by determining the students’
brain dominance level and track their learning progress based
on the EEG signal in a scientific approach.
Index Terms—brain dominance, correlation,
electroencephalogram, topographical, power spectral density
I. INTRODUCTION
HE brain is known as the most complex organ in the
human body. The brain basically can be split into two
parts, which are the right hemisphere and the left hemisphere.
The brain is an unpredictable organ which is also in charge of
learning, sensing, body controlling, and memory [1]. The left
and right hemispheres are interconnected by the corpus
callosum, which is located in the middle between the left and
right hemispheres. It enables the information from the left
hemisphere to flow to the right hemisphere and vice versa.
Manuscript received May 22, 2020; This work was supported in part by
the Telekom Malaysia Research & Development Sdn Bhd grant (MMUE/190088).
Z. Y. Lim is with the Faculty of Engineering and Technology, Multimedia
University, Jalan Ayer Keroh Lama, 75450 Melaka (phone: +60143471296; e-mail: limzhengyou@gmail.com).
Both the left hemisphere and right hemisphere serve different
functions [2]. The left half movements of the body are in
charged by the right hemisphere and vice versa. Since 1981,
Roger Sperry found out that both brain hemispheres could
carry out different functions. The left hemisphere is majoring
in sequential thinking, logical thinking, mathematical
problem solving, and analysis. Whilst, the right brain is
majoring in creative thinking, imagination, music, art, and
emotions [3]. Most people tend to dominantly use one side of
the brain, thus usually people are to be determined either left
or right brain dominance [4]. A left brain thinker who is left-
brain dominant tends to have a better talent in studying
science and mathematics that involves analytical and logical
thinking. Contrarily, a right brain thinker who is right-brain
dominant tends to have a greater ability in art and music that
requires a higher level of imagination and creativity.
However, a person can also maximize the usage of both
sides of the hemisphere. Thus, the person is known as left and
right brain balanced and also called as the whole brain thinker
[4]. Several researches show that achieving brain balancing
could unleash the potential of brain capability and it is one of
the main reasons which leads to successful achievements in
one person’s life [5]. A few greatest representatives who are
whole-brain thinkers are Albert Einstein, Leonardo Da Vinci,
Samuel Morse, etc. They made high achievements in
scientific research or inventions, along with high attainment
in art or music. It is shown that imagination and creativity
thinking majoring in the right brain are as important as the
scientific and logical thinking majoring in the left brain
[6][7]. Besides, in the field of education, researches also show
that achieving brain synchronization could aid students in
rapid learning. Thus, some popular academicians have
innovated education system which can develop the right
hemisphere of the brain such as Betty Edwards and Makoto
Shicida. Their students’ achievements prove the results of
their new education concept and approach.
To determine one person’s brain dominance level, the
Stroop’s test and questionnaire-based approach such as the
Herrmann Brain Dominance Instrument are employed [8].
Nevertheless, currently there is still no data-driven approach
to validate the brain dominance test. There are a few
approaches that can acquire the information of the brain
activity: positron emission tomography (PET), magnetic
resonance imaging (MRI), functional magnetic resonance
imaging (fMRI), and electroencephalogram (EEG) [9].
K.S. Sim is with the Faculty of Engineering and Technology, Multimedia
University, Jalan Ayer Keroh Lama, 75450 Melaka (e-mail: kssim@mmu.edu.my).
S. C. Tan is with the Faculty of Information Science and Technology,
Multimedia University, Jalan Ayer Keroh Lama, 75450 Melaka (e-mail: sctan@mmu.edu.my).
An Evaluation of Left and Right Brain
Dominance using Electroencephalogram Signal
Z. Y. Lim, K. S. Sim, and S. C. Tan
T
Engineering Letters, 28:4, EL_28_4_46
Volume 28, Issue 4: December 2020
______________________________________________________________________________________
-
Among these methods, EEG is the most suitable method as it
is non-invasive, portable, and high sampling frequency [10].
EEG measures the voltage emitted from the scalp due to
the ionic current flows within the neurons in the brain. The
voltage obtained from the scalp is around 10 microvolts to
100 microvolts [11]. Thus, a signal amplifier is required to
amplify the acquired signal. Besides, the raw EEG waveform
is complicated and may be contaminated by different sources
of noise. Thus, it requires preprocessing techniques such as
baseline removal [12], independent component analysis
(ICA) [13], etc. to remove the noise and artifacts from the
acquired raw EEG signal [14][15]. In order to analyze the
complicated time-based waveform, EEG Topographical
Power Spectral Density Percentage (EEGTPSDP) method is
implemented based on several techniques such as frequency-
based [16][17], power spectral density [18][19], etc. are
employed. In this paper, the process involved in the EEG data
acquisition, preprocessing techniques, analysis method, and
the correlation between the brain dominance level and EEG
signal are presented.
II. IMPLEMENTATION
A. Electroencephalogram Signal Acquisition System
The EEG device employed in this research is known as the
Ultracortex “Mark IV” EEG Headset developed by OpenBCI
as shown in Fig. 1.
Fig. 1. Open BCI Mark IV EEG headset
It is a 3D printed headset equipped with Cyton Biosensing
board which allows the acquired data to transmit wirelessly
from the EEG device to the personal computer. The
advantage of the headset is that it allows the user to configure
the location of the electrode sensors among the 35 node
locations as shown in Fig. 2, where the figure is redrawn
based on the image indicates the node locations of Ultracortex
Mark IV in OpenBCI shop [20].
The 35 node locations are designed according to the
International 10/20 system [21]. The numbers ‘10’ and ‘20’
represents the distances between every position of the
sensors. They are either 10% or 20% where the total length is
based on the length from the front to the back of the head, or
left side to the right side of the head. The electrode positions
are represented by letter: F, T, C, P and O. ‘F’ indicates the
frontal lobe, ‘T’ indicates the temporal lobe, ‘C’ indicates
central, ‘P’ indicates the parietal lobe, and O indicates
“Occipital lobe”. The letter will be followed by either letter
‘z’ or numbers. ‘z’ indicates zero and refers to the location
on the centerline which separates the left and right
Fig. 2. 35 node locations for positioning electrode sensors (Redrawn based
on the image indicates the node locations of Ultracortex Mark IV in
OpenBCI shop) [20]
hemisphere. Even numbers refer to the location on the right
hemisphere and odd numbers refer to the location on the left
hemisphere. Fig. 3 shows the illustration of the International
10/20 system, which is redrawn based on figures and
information on page 140 of “Electroencephalography: Basic
Principles, Clinical Applications, and Related Fields” by E.
Niedermeyer and F. L. d. Silva. [21].
Fig. 3. International 10/20 system (Redrawn based on figures and
information in page 140 of “Electroencephalography: Basic Principles,
Clinical Applications, and Related Fields” by E. Niedermeyer and F. L. d.
Silva) [21]
In order to determine the signal imbalance level between
the left and right hemispheres of the brain, the number and
position of electrodes should be equal and symmetry for both
left and right hemispheres. The frontal lobe of the brain
majoring is responsible for concentration and solving
complex problems, thus sensor electrodes are placed at the
position of Fp1 and Fp2. Since the parietal lobe of the brain
majoring is in the task management and working memory,
hence sensor electrodes are fixed at the position of F7 and F8.
The temporal lobe is responsible for hearing and long term
memory, thus two sensor electrodes are placed at the position
of P7 and P8. The occipital lobe of the brain mainly functions
for vision and sight, thus electrodes are placed at the two
positions of O1 and O2.
Engineering Letters, 28:4, EL_28_4_46
Volume 28, Issue 4: December 2020
______________________________________________________________________________________
-
B. EEG Signal Acquisition Procedure
In this research, there are a total number of 60 samples. The
subjects are required to wear the EEG device throughout the
signal acquisition process. The procedure of the EEG signal
acquisition process is illustrated in Fig. 4.
Fig. 4. Procedure of the EEG signal acquisition process
In the beginning, the EEG device is placed onto the
subject’s head and the contact quality of all the sensors are
ensured in good condition. Then, the subject is requested to
rest calmly with opening the eyes for 2 minutes when the
EEG signals are recorded. After that, the subject is asked to
perform the Hermann Brain Dominance Instrument test with
a personal computer. The active state EEG of the subject is
recorded during the subject performing the test. After the
subject has done the test, the subject is requested to rest and
relax with opening the eyes again for another 2 minutes.
Meanwhile, the post resting state EEG signal is recorded.
C. Hermann Brain Dominance Instrument
Hermann Brain Dominance Instrument is a psychometric
assessment developed by William Hermann to determine the
strength of each cognitive style represented by the left and
right hemispheres of the brain [8]. This test is widely
employed by many companies and employers to determine
the cognitive ability and personalities of the employees. The
concept of HBDI is thinking and can be categorized into four
modes which are analytical thinking, sequential thinking,
interpersonal thinking, and imaginative thinking. The HBDI
test requires the user to answer the questionnaires designed to
measure the degree of preference for each of the four modes
of thinking. Fig. 5 shows the whole brain model concept of
HBDI, the figure is redrawn based on the information of
Whole Brain® Model in Think Hermann-How it works [8].
Fig. 5. Whole Brain Model concept of HBDI (Redrawn based on the
information of Whole Brain® Model in Think Hermann-How it works) [8]
The results of the test will show the strength of the subject
in each quadrant. The quadrant A represents the front left side
of the brain indicates the person’s strength in analytical
thinking which includes logical thinking, mathematics, and
evidence-based decision making. The quadrant B represents
the rear left side of the brain indicates the person’s ability in
sequential thinking which includes detailed planning, timing,
and scheduling. The quadrant C represents the rear right side
of the brain indicates emotional thinking and interpersonal
thinking which involves communication and feeling. And the
quadrant D represents the front right side indicates the
person’s imaginative thinking and creativity. According to
the results of the test, the strength percentage of each quadrant
can figure out the person is left-brain dominance, right-brain
dominance, or whole-brain thinker and how balance is their
left and right brain hemisphere.
D. EEG Signal Preprocessing
As the original voltage of the EEG signal is extremely low
and digital amplifier is employed, the EEG signal is often
contaminated by different sources of noise such as electrical
line noise, muscular activity, eye blinking, and so on. Thus,
signal preprocessing is mandatory to eliminate the
unwanted artifacts from the acquired raw signal. Fig. 5
shows the preprocessing techniques to be employed in this
research.
Fig. 6. Block Diagram of EEG Signal Preprocessing Techniques
Generally, the EEG waveform does not oscillate at a
baseline voltage such as alternating current (AC) electrical
waveform, which oscillates at the voltage value of zero
[22]. The baseline voltage where the EEG waveform
oscillates will shift time by time as shown in Fig.7., where
the black line indicates the baseline of the waveform.
Fig.7. Raw EEG Waveform Plot
The baseline correction is performed by using the equation
as shown in (1) to (5).
𝑓𝑛[𝑥] = {
𝑓𝑛𝑅[𝑥] + 𝐶𝑛
𝐿 , 0 ≤ 𝑥 < 𝑊𝑛 − 11
2∗ (𝑓𝑛
𝑅[𝑥] + 𝑓𝑛𝐿[𝑥]) , 𝑊𝑛 − 1 ≤ 𝑥 ≤
𝑓𝑛𝐿[𝑥] + 𝐶𝑛
𝑅 , 𝑘 − 𝑊𝑛 < 𝑥 ≤ 𝑘 − 1
𝑘 − 𝑊𝑛 (1)
Engineering Letters, 28:4, EL_28_4_46
Volume 28, Issue 4: December 2020
______________________________________________________________________________________
-
where
𝑓𝑛𝐿[𝑥] =
1
𝑊𝑛∑ 𝑓𝑛−1[𝑥] , 𝑊𝑛 − 1 ≤ 𝑥 ≤ 𝑘 − 1
𝑥𝑣=𝑥−𝑊𝑛+1
(2)
𝑓𝑛𝐿[𝑥] =
1
𝑊𝑛∑ 𝑓𝑛−1[𝑥] , 0 ≤ 𝑥 ≤ 𝑘 − 𝑊𝑛
𝑥+𝑊𝑛−1𝑣=𝑥 (3)
𝐶𝑛𝐿 =
1
2[[𝑓𝑛
𝐿[𝑊𝑛 − 1] − 𝑓𝑛𝐿[𝑊𝑛 − 1]] (4)
𝐶𝑛𝑅 =
1
2[[𝑓𝑛
𝑅[𝑘 − 1] − 𝑓𝑛𝐿[𝑘 − 𝑊𝑛]] (5)
where 𝑓𝑛[𝑥] is the output value, 𝑓𝑛−1[𝑥] is the input value, 𝑓𝑛
𝑅[𝑥] is right moving average, 𝑓𝑛𝐿[𝑥] is left moving average,
𝐶𝑛𝐿 is compensation value for the left moving average, 𝐶𝑛
𝑅 is
compensation value for right moving average, 𝑊𝑛 is window length, 𝑛 is the number of order for moving average filter.
Next, the electrical line noise is removed by using a notch
filter as implemented using (6).
𝐺(𝑧) =1−2 cos(50)𝑧−1+𝑧−2
1−2𝑟𝑐𝑜𝑠(50)𝑧−1+𝑟2𝑧−2 (6)
where 𝑧 is the z-transform of the signal. Next, independent component analysis is employed to
eliminate the unwanted noise caused by eye blinking,
muscular action, etc. The input signal is modeled as (7).
𝑌 = 𝐵 ∙ 𝑆 (7) where 𝑌 is matrix represents input signal, 𝐵 is the matrix that represents source values and 𝑆 is the matrix to represents the number of sources. Equations (8) to (13) are used to estimate
out the value of the sources.
𝐵 = 𝑃𝛾−1𝑄𝑇 (8) where
𝑄𝑇 = [𝑐𝑜𝑠𝜃 𝑠𝑖𝑛𝜃
−𝑠𝑖𝑛𝜃 𝑐𝑜𝑠𝜃] (9)
𝛿1 = ∑ 𝑌(𝑗)𝑐𝑜𝑠𝜃𝑁𝑗=1 (10)
𝛿2 = ∑ 𝑌(𝑗)cos (𝜃 +𝜋
2)𝑁𝑗=1 (11)
𝛾−1 = [
1
√𝛿10
01
√𝛿2
] (12)
𝑃 = [𝑐𝑜𝑠∅ 𝑠𝑖𝑛∅
−𝑠𝑖𝑛∅ 𝑐𝑜𝑠∅] (13)
where 𝑁 is the number of sources, 𝜃 is the angle of rotation matrix and ∅ is the angle value at the minimum value of normalized kurtosis. Then, the final reconstructed signal is
expressed as (14).
𝑆 =̃ [𝑐𝑜𝑠𝜃 𝑠𝑖𝑛𝜃
−𝑠𝑖𝑛𝜃 𝑐𝑜𝑠𝜃] ∙ [
1
√𝛿10
01
√𝛿2
] ∙ [𝑐𝑜𝑠∅ 𝑠𝑖𝑛∅
−𝑠𝑖𝑛∅ 𝑐𝑜𝑠∅] (14)
E. EEG Topographical Power Spectral Density Percentage (EEGTPSDP)
As time-based EEG waveform is complicated and
significant features are unable to be identified from the
complicated time-based EEG waveform, several processing
techniques are applied to the EEG waveform to ease the
process of analysis and feature extraction. The first technique
employed is the Continuous Wavelet Transform (CWT) that
converts the complicated time-based waveform into a
frequency-based waveform. The technique is implemented
by using equation (15).
𝐹(𝑥, 𝑦) =1
√|𝑥|∫ 𝑓(𝑡)𝜑 (
𝑡−𝑦
𝑥) 𝑑𝑡
+∞
−∞ (15)
where φ(t) is known as the defined wavelet function, 𝑥 is the
dilation scaling parameter and 𝑦 is the translation scaling parameter. In this way, the time-based waveform will be
converted into a frequency-based waveform.
Next, power spectral density is obtained based on the
frequency-based waveform. It is obtained by using equation
(16) and (17).
𝑃 = ∫ |�̃�(𝑓)|2𝑑𝑓
+∞
−∞ (16)
�̃�(𝑓) = ∫ 𝑒2𝜋𝑓𝑡𝑥(𝑡)2𝑑𝑡∞
−∞ (17)
where 𝑓 is frequency-based and 𝑡 is time-based. Based on the results of the power spectral density, a 2D
topographical interpolation plot can be constructed by
defining a set of grid points based on the 10/20 system with
the value of power spectral density. Fig 8 shows the grid
points defined on the 2D head plot top view.
Fig. 8 Grid points defined on the 2D head plot (top view)
Color intensity is used to represent and differentiate the
correlative strength of the power spectral density. The radius
of the color which represents the maximum value is computed
based on the normalized value of the power spectral density
and predefined maximum radius size as shown in equation
(18).
𝑅𝑛 =max (�̃�𝑛(𝑓))
max (∀�̃�𝑛(𝑓))∗ 𝑟𝑠 (18)
where 𝑅𝑛 is the radius size for the maximum intensity, max (�̃�𝑛(𝑓) is the maximum value of power spectral intensity for the channel 𝑛, where 𝑛 ∈ (1,2,3,4,5,6,7,8) and 𝑟𝑠 is the predefined maximum radius size for each grid, wherein this case is 6 cm. Except for the color that represents
the maximum intensity, the other colors are diffuse evenly
until reaching the predefined maximum size. After that, color
interpolation is employed in between the grids where they
intersect to smoothen the topographical view.
Next, the EEG Topographical Power Spectral Density
Percentage (EEGTPSDP) value is computed according to the
resultant interpolated intensity value on the topographical
plot. The value indicates the percentage of power spectral
density which is contributed by each hemisphere of the brain.
The EEGTPSDP value for the left and right hemisphere are
calculated using (19) and (20).
𝐸𝐸𝐺𝑇𝑃𝑆𝐷𝑃𝐿 =(2 ∑ 𝑅𝑙)+∑ 𝐵𝑙
(2 ∑ 𝑅𝑤)+∑ 𝐵𝑤∗ 100% (19)
𝐸𝐸𝐺𝑇𝑃𝑆𝐷𝑃𝑅 =(2 ∑ 𝑅𝑟)+∑ 𝐵𝑟
(2 ∑ 𝑅𝑤)+∑ 𝐵𝑤∗ 100% (20)
where ∑ 𝑅𝑙 is the total value of red pixels on the left hemisphere, ∑ 𝐵𝑙 is the total value of blue pixels on the left hemisphere, ∑ 𝑅𝑟 is the total value of red pixels on the right hemisphere, ∑ 𝐵𝑟 is the total value of blue pixels on the right hemisphere, ∑ 𝑅𝑤 is the total value of red pixels on the whole brain topography, and ∑ 𝐵𝑤 is the total value of blue pixels on the whole brain topography.
Engineering Letters, 28:4, EL_28_4_46
Volume 28, Issue 4: December 2020
______________________________________________________________________________________
-
III. RESULTS AND DISCUSSION
A. Frequency-based analysis
By using the CWT, a 3-axes time-frequency plot is
constructed based on the time-based waveform. The time-
frequency plot represents the strength of the respective
frequency against time by using color intensity. Fig. 9 shows
the example of the generated time-frequency plot from
channel 1 to channel 8 from a left-brain dominant subject.
(a)
(b)
(c)
(d)
As shown in Fig.3, the odd number channels are located on
the left hemisphere and the even number channels are located
(e)
(f)
(g)
(h)
Fig. 9 Time-frequency plot for (a) channel 1, FP1 (b) channel 2, FP2 (c)
channel 3. F7 (d) channel 4, F8 (e) channel 5, P7 (f) channel 6, P8 (g)
channel 7, O1 (h) channel 8, O2
on the right hemisphere. The position of the odd number
channel is symmetrical to the next even number. For instance,
the position of channel 1 (FP1) on the left hemisphere is
symmetrical to the position of channel 2 (FP2) on the right
hemisphere, position of channel 3 (F7) on the left hemisphere
is symmetrical to the position of channel 4 (F8), and so on.
Hence, the results of the channels can be compared in pairs
(e.g. channel 1 compared to channel 2, channel 3 compared
Engineering Letters, 28:4, EL_28_4_46
Volume 28, Issue 4: December 2020
______________________________________________________________________________________
-
to channel 4, and so on) to determine the difference between
the left and right hemispheres of the brain.
The results generated and shown in Fig.9(a) to Fig.9(h) are
from a subject who is a left-brain dominant. The time-
frequency plots show a major difference where the odd
channels have a larger distribution area of high amplitudes
signal in the range of 10Hz to 40Hz compared to channel 2.
This indicates the region on the left hemisphere is more active
compared to the right hemisphere.
Next, in order to analyze in terms of the average power in
each frequency, the power spectral density (PSD) graph is
generated according to the value from the time-frequency
plot. Fig.10 shows the generated power spectral density graph
for channel 1 and channel 2.
(a)
(b)
(c)
(d)
From the generated results shown in Figure 10, they show
that the PSD values of odd channels (Channel 1, 3, 5, and 7)
(e)
(f)
(g)
(h)
Fig. 10 Power spectral density chart for (a) channel 1, FP1 (b) channel 2,
FP2 (c) channel 3. F7 (d) channel 4, F8 (e) channel 5, P7 (f) channel 6, P8
(g) channel 7, O1 (h) channel 8, O2
are higher compared to even channels (Channel 2, 4, 6, and
8) especially in the range of 3Hz to 40Hz. This indicates that
the average power spectral density in every frequency is
higher in the dominant brain hemisphere in contrast with
another brain hemisphere. Hence, results show that the left
brain dominant person tends to have a higher average PSD
value on the left brain hemisphere, whereas the right brain
dominant person tends to have a higher average PSD value
on the right brain hemisphere. As the EEG signal is
commonly analyzed in term of the frequency band, the data
from the PSD chart is quantified according to the frequency
bands: Delta (0.5Hz to 4Hz), Theta (4Hz to 8Hz), Alpha (8Hz
to 13Hz), Beta (13Hz to 30Hz) and Gamma (above 30Hz).
Figure 11 shows the average power spectral density in each
Engineering Letters, 28:4, EL_28_4_46
Volume 28, Issue 4: December 2020
______________________________________________________________________________________
-
frequency band for 8 channels.
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
Fig. 11 Average power spectral density in frequency band chart for (a)
channel 1, FP1 (b) channel 2, FP2 (c) channel 3. F7 (d) channel 4, F8 (e)
channel 5, P7 (f) channel 6, P8 (g) channel 7, O1 (h) channel 8, O2
Engineering Letters, 28:4, EL_28_4_46
Volume 28, Issue 4: December 2020
______________________________________________________________________________________
-
From the results shown in Figure 11 in terms of the
frequency band, they show a significant difference between
the left hemisphere (Channel 1, 3, 5, and 7) and right
hemisphere (Channel 2, 4, 6, and 8) especially in the Theta
(from 4Hz to 8Hz), Beta frequency band (from 13Hz to 30Hz)
and Gamma frequency band (above 30Hz). The EEG signal
is collected during resting state with opening eyes. Theta
frequency band indicates the resting state, the Beta frequency
band indicates the brain is in an active state with eyes opened,
and the Gamma frequency band indicates brain activity of
information processing such as memory, thinking, and
consciousness. The results show that the PSD value is higher
on the left hemisphere (odd number channels) compared to
the right hemisphere (even number channels) in the Theta,
Beta, and Gamma frequency band. Hence, the dominant brain
hemisphere of a person can be identified through the
comparison of PSD values in the Theta, Beta, and Gamma
frequency band between the two hemispheres of the brain.
Although the features of left or right brain dominance can
be identified from the comparison of the time-frequency plot
and power spectral density chart between the left and right
channels, but the process of comparing the charts is tedious.
Thus, the topographical view constructed based on the power
spectral density can provide better visualization and obvious
results.
B. Topographical plot
According to the power spectral density of all the 8
channels, topographical plots are constructed according to the
PSD and predefined position of channels on the 2D head plot.
Fig.12 shows the topographical plot.
Fig. 12 Topographical plot based on power spectral density
As shown in Fig.12, the topographical plot is able to
visualize and compare the PSD which is contributed by a
different region of the brain. Thus, the power intensity of the
topographical plot is compared with the results obtained
through HBDI assessment. Fig. 13 shows some of the
topographical results and the left and right hemispheres are
separated by the middle dashed line.
Fig. 13 Topographical plot of (a) subject with left-brain dominant (b)
subject with right-brain dominant (c) subject with left-brain dominant (d)
subject with right-brain dominant
From the observation in Fig. 13, it shows that the
topographical plot based on PSD contributed by the eight
channels tends to have higher intensity on the left hemisphere
for a left-brain dominant person, and vice versa.
C. Brain Dominance Level vs EEGTPSDP Value
As all the subjects are required to undergo the HBDI
assessment, the brain dominance level of each brain
hemisphere can be obtained through the results of the
assessment. The HBDI results act as benchmark data to
classify the person is either left- or right-brain dominant.
The brain dominance level results according to the HBDI
assessment carried out by the subjects and the respective
EEGTPSDP value are computed and tabulated. 18 of the
samples are tabulated in the TABLE I and illustrated in
scatter plots as shown in Fig. 14 and Fig.15.
By compiling the EEGTPSDP value according to the
HBDI results and plotted in the scatter graphs as shown in
Fig.14 and Fig.15,
TABLE I
HBDI ASSESSMENT RESULTS AND POWER SPECTRAL DENSITY FOR LEFT
AND RIGHT HEMISPHERE OF BRAIN
Fig.14 Scatter Plot of EEGTPSDP vs HBDI Assessment Result in
Percentage For Left Hemisphere
Left Right Left Right
1 11.11 88.89 30 70
2 60.71 39.29 55 45
3 30 70 20 80
4 55.56 44.44 80 20
5 60.71 39.29 90 10
6 20 80 40 60
7 44.44 55.56 25 75
8 53.57 46.43 60 40
9 40 60 40 60
10 61.11 38.89 85 15
11 64.29 35.71 65 35
12 50 50 60 40
13 55.56 44.44 85 15
14 53.57 46.43 75 25
15 50 50 45 55
16 44.44 55.56 40 60
17 39.29 60.71 20 80
18 60 40 70 30
HBDI Results (%) EEGTPSDP (%)Sample
Engineering Letters, 28:4, EL_28_4_46
Volume 28, Issue 4: December 2020
______________________________________________________________________________________
-
Fig.15 Scatter Plot of EEGTPSDP vs HBDI Assessment Result in
Percentage for Right Hemisphere
The simple linear regression line (red line) computed based
on the scatter graphs show that the relationship between brain
dominance level and EEGTPSDP value is proportional. This
indicates that the higher percentage of the left brain
dominance level (based on HBDI assessment), the higher the
power spectral density is on the left hemisphere and vice
versa.
D. Correlation Coefficient of Brain Dominance Level vs EEGTPSDP Value
In order to determine the correlation between the two
variables, few correlation coefficients are employed. The first
one is the Pearson correlation coefficient, which is calculated
based on (21) and (22).
𝑟𝐿 =𝑛(∑ 𝑥𝐿∗𝑦𝐿)−(∑ 𝑥𝐿)∗(∑ 𝑦𝐿)
√[(𝑛 ∑ 𝑥𝐿2−(∑ 𝑥𝐿)
2)∗(𝑛 ∑ 𝑦𝐿
2−(∑ 𝑦𝐿)2
)
(21)
𝑟𝑅 =𝑛(∑ 𝑥𝑅∗𝑦𝑅)−(∑ 𝑥𝑅)∗(∑ 𝑦𝑅)
√[(𝑛 ∑ 𝑥𝑅2 −(∑ 𝑥𝑅)
2)∗(𝑛 ∑ 𝑦𝑅
2−(∑ 𝑦𝑅)2
)
(22)
where 𝑟𝐿 is the Pearson correlation coefficient for the left hemisphere, 𝑟𝑅 is the Pearson correlation coefficient for the right hemisphere, 𝑛 is the total number of sample, 𝑥𝐿 is the left hemisphere dominance level according to HBDI results,
𝑦𝐿 is the EEGTPSDP value for the left hemisphere, 𝑥𝑅 is the right hemisphere dominance level according to HBDI results,
𝑦𝑅 is the EEGTPSDP value for the right hemisphere. For the Pearson correlation coefficient, -1.0 indicates
linearly inverse proportional, 1.0 indicates linearly
proportional and 0 indicates no correlation [23]. As the
computed value for 𝑟𝐿 and 𝑟𝑅 are both 0.73, it shows that both the variables which are the HBDI result and power spectral
density have a significant proportional relationship.
Another correlation method, namely Spearman rank
correlation, is employed to determine the correlation between
the EEG PSD percentage and brain dominance level. The
coefficient is calculated using (23) and (24).
𝜌𝐿 =1
𝑛∑ [𝑅(𝑥𝐿𝑖)−𝑅(𝑥𝐿)
̅̅ ̅̅ ̅̅ ̅̅ ]∙[𝑅(𝑦𝐿𝑖)−𝑅(𝑦𝐿)̅̅ ̅̅ ̅̅ ̅̅ ]𝑛𝑖=1
√(1
𝑛∑ [𝑅(𝑥𝐿𝑖)−𝑅(𝑥𝐿)
̅̅ ̅̅ ̅̅ ̅̅ ]2)∙(1
𝑛∑ [𝑅(𝑥𝐿𝑖)−𝑅(𝑥𝐿)
̅̅ ̅̅ ̅̅ ̅̅ ]2)𝑛𝑖=1𝑛𝑖=1
(23)
𝜌𝑅 =1
𝑛∑ [𝑅(𝑥𝑅𝑖)−𝑅(𝑥𝑅)
̅̅ ̅̅ ̅̅ ̅̅ ̅]∙[𝑅(𝑦𝑅𝑖)−𝑅(𝑦𝑅)̅̅ ̅̅ ̅̅ ̅̅ ̅]𝑛𝑖=1
√(1
𝑛∑ [𝑅(𝑥𝑅𝑖)−𝑅(𝑥𝑅)
̅̅ ̅̅ ̅̅ ̅̅ ̅]2)∙(1
𝑛∑ [𝑅(𝑦𝑅𝑖)−𝑅(𝑦𝑅)
̅̅ ̅̅ ̅̅ ̅̅ ̅]2)𝑛𝑖=1𝑛𝑖=1
(24)
where 𝜌𝐿 is the Spearman rank correlation coefficient for the left hemisphere, 𝜌𝑅 is the Spearman rank correlation coefficient for the right hemisphere, 𝑅(𝑥𝐿) is the rank of the
variable 𝑥𝐿 , and 𝑅(𝑥𝐿)̅̅ ̅̅ ̅̅ ̅̅ is the mean rank of the variable 𝑥𝐿. Spearman rank correlation coefficient is similar to the
Pearson correlation coefficient, where -1.0 indicates a perfect
negative correlation, 1.0 indicates a perfect positive
correlation and 0 indicates no correlation [23]. The computed
value for 𝜌𝐿 and 𝜌𝑅 are both 0.82, it shows that both the EEGTPSDP value in one hemisphere and brain dominance
have a significant proportional relationship.
The third correlation coefficient method employed to
assess the relationship between EEGTPSDP value and brain
dominance level is known as Kendall’s Tau correlation
coefficient. They are calculated using (25) and (26).
𝜏𝐿 =(𝐶𝐿−𝐷𝐿)
(𝐶𝐿+𝐷𝐿) (25)
𝜏𝑅 =(𝐶𝑅−𝐷𝑅)
(𝐶𝑅+𝐷𝑅) (26)
where 𝜏𝐿 is the Kendall’s Tau correlation coefficient for the left brain hemisphere, 𝐶𝐿 is the number of concordant pairs for the left brain hemisphere, 𝐷𝐿 is the number of discordant
pairs for the left brain hemisphere, 𝜏𝑅 is the Kendall’s Tau correlation coefficient for the right brain hemisphere, 𝐶𝑅 is the number of concordant pairs for the right brain
hemisphere, 𝐷𝑅 is the discordant pairs for the right brain hemisphere. Concordant pair is defined as the pair of
variables with proportional relationship and discordant pair is
defined as the pair of variables with an inversely proportional
relationship.
The computed value for 𝜏𝐿 and 𝜏𝑅 are both 0.97. For Kendall’s Tau correlation coefficient, 0 indicates no
relationship, and 1.0 indicates a perfect relationship [23].
Hence, this shows the EEGTPSDP value in one hemisphere
is highly correlated with the dominant brain hemisphere. The
computed values of all the three correlation coefficients are
summarized in Table II.
TABLE II CORRELATION COEFFICIENTS OF BRAIN DOMINANCE LEVEL VS
EEGTPSDP VALUE
Correlation
coefficient
Left hemisphere Right hemisphere
Pearson 0.73 0.73
Spearman Rank 0.82 0.82
Kendall’s Tau 0.97 0.97
According to the computed results summarized in Table II,
where Pearson correlation coefficient value is 0.73, Spearman
rank correlation coefficient value is 0.82, and Kendall’s Tau
correlation coefficient value is 0.97. This validates that the
dominant brain hemisphere has a very strong correlation with
EEGTPSDP value which is computed from the EEG power
spectral density generated by the respective hemisphere. A
left-brain dominant person will have a higher EEGTPSDP
value on the left hemisphere, whereas a right-brain dominant
person will have a higher EEGTPSDP value on the right
hemisphere. This also proves that the EEG signal can be
utilized to determine one person is either left-brain dominant
or right-brain dominant in a scientific manner.
IV. CONCLUSION
In this research, we have implemented several EEG signal
preprocessing techniques to eliminate the noise and artifacts
from the acquired signal. Besides, the analysis technique
implemented in this research namely EEGTPSDP is an
efficient method to classify brain dominance. Based on the
Engineering Letters, 28:4, EL_28_4_46
Volume 28, Issue 4: December 2020
______________________________________________________________________________________
-
results of the HBDI assessment carried out by the subject and
the computed EEGTPSDP value, it can be concluded that the
brain dominance level has a very high correlation with the
EEG power spectral density contributed by the respective
hemisphere. Thus, this proves that the acquired EEG signal
along with the EEGTPSDP method can be used to classify
whether the person is left- or right-brain dominant.
By employing the EEG analysis method which reflects the
biofeedback information from the brain, this research proves
that the left and right dominance theory is no longer a myth.
It is believed that the implemented system and the findings in
this research are contributing to the first step in
revolutionizing the current one-for-all education system. By
knowing the brain dominance of the student, the teachers can
design appropriate syllabus and training according to their
students’ brain dominance level. In this way, it will be able to
unleash the full brain potential and ability of every student.
ACKNOWLEDGMENT
In this research, we would like to express our gratitude to
the subjects who are volunteered to participate in the HBDI
assessment and EEG data collection.
REFERENCES
[1] M. Oflaz, "The effect of right and left brain dominance in language learning", Procedia Social and Behavioral Sciences, vol.15, pp.1507-1513, 2020.
[2] S. A. Ali and S. Raza, "A study of right and left brain dominant students at IB&M with respect to their gender, age and educational background," International Journal of Advances in Scientific Research,
vol. 3(9), pp. 115-120, 2017.
[3] A. Trafton, “Synchronized brain waves enable rapid learning,” MIT News, 12 June 2014, [Online]. Available:
https://news.mit.edu/2014/synchronized-brain-waves-enable-rapid-
learning-0612. [Accessed 6 May 2020]. [4] D. Deardorff, " An exploratory case study of leadership influences on
innovative culture: A descriptive study,” Dissertation Abstracts
International: Section B: The Sciences and Engineering, vol. 66(4B), pp.2338, 2005.
[5] H. Ashraf, T. Branch and M. T. Yazdi, “Brain Dominance Quadrants and Reflective Teaching among ELT Teachers: A Relationship Study”,
International Journal of English Linguistics, vol. 7(2), pp. 63-72, 2017.
[6] H. Kwon, J. Cho, E. Lee, “EEG Asymmetry Analysis of the Left and Right Brain Activities During Simple versus Complex Arithmetic
Learning,” Journal of Neurotherapy, vol. 13, pp.109-116, 2009.
[7] K. Cherry, “Left Brain vs. Right Brain Dominance,” Very Well Mind, 10 April 2020, [Online]. Available:
https://www.verywellmind.com/left-brain-vs-right-brain-2795005.
[Accessed 6 May 2020] [8] “Herrmann's Core Idea: Whole Brain® Thinking," Think Hermann,
[Online]. Available: https://www.thinkherrmann.com/how-it-works/.
[Accessed 6 May 2020] [9] K.S Sim, Z.Y Lim, T.K. Kho, "Brainwave Controlled Electrical
Wheelchair” International Conference on Automations Sciences 2016,
vol.1, pp. 10-15, 2016. [10] K. S. Sim, Z. Y. Lim, S. Fawaz, "Development of Dementia
Neurofeedback System using EEG Brainwave Signals", International
Journal of Signal Processing Systems, vol.7, pp. 113-117, 2019. [11] K.S Sim, Z.Y Lim, T.K. Kho, “EEG Controlled Wheelchair”, MATEC
Web of Conferences, vol. 51, pp 18-25, 2016.
[12] P. M. Alday, "How much baseline correction do we need in ERP research? Extended GLM model can replace baseline correction while
lifting its limits ," Psychophysiology , vol. 56 (12), pp. 1-22, 2019.
[13] A. Tharwat, “Independent component analysis: An introduction”, Applied Computing and Informatics, 2018.
[14] S. Leske and S. S. Dalal, "Reducing power line noise in EEG and MEG data via spectrum interpolation, vol. 189, pp. 763-776, 2019.
[15] S. S. Daud and R. Sudirman, "Butterworth Bandpass and Stationary Wavelet Transform Filter Comparison for Electroencephalography
Signal ," 6th International Conference on Intelligent Systems, Modelling and Simulation, pp.123-126, 2015.
[16] A. S. Al-Fahoum and A. A. Al-Fraihat, “Methods of EEG Signal Features Extraction Using Linear Analysis in Frequency and Time-
Frequency Domains”, ISRN Neuroscience, 2014.
[17] K. S. Sim, Z. Y. Lim, “Fast Fourier Analysis and EEG Classification Brainwave Controlled Wheelchair”, 2nd International Conference on
Control Science and Systems Engineering (ICCSSE 2016), pp. 12-17,
2016. [18] R. Wang, J. Wang, H. Yu, et al., "Power spectral density and coherence
analysis of Alzheimer's EEG.," Cogn Neurodyn, vol. 9(3), pp 291-304,
2015. [19] N. P. Subramaniyam, “Measuring Entropy in the EEG,” Sapien Labs,
5 Feb 2018, [Online]. Available: https://sapienlabs.org/measuring-
entropy-in-the-eeg/. [Accessed on 6 May 2020] [20] “Ultracortex Mark IV EEG Headset”, OpenBCI Online Store, [Online].
Available: https://shop.openbci.com/products/ultracortex-mark-iv.
[Accessed 29 Sep 2020] [21] E. Niedermeyer and F. L. d. Silva, Electroencephalography: Basic
Principles, Clinical Applications, and Related Fields, Lippincott
Williams & Wilkins, pp. 140, 2004. [22] S. Fawaz, K. S. Sim and S. C. Tan, "Encoding Rich Frequencies for
Classification of Stroke Patients EEG Signals," IEEE Access, vol. 8,
pp. 135811-135820, 2020, [23] U, R. Sudha, V. Ragavi and C. Thirumalai, "Analyzing correlation
coefficient using software metrics," International Conference on
Trends in Electronics and Informatics (ICEI), pp. 1151-1153, 2017.
Z. Y. Lim received the B.S. degree in electronic engineering majoring in
robotics and automation from Multimedia University Malaysia, in 2016 and
pursuing the M.S. degree in Engineering Science from Multimedia University Malaysia. He is currently a PhD Research Scholar in Multimedia
University since Jan 2020. His research interest is in electroencephalogram
(EEG), artificial intelligence, robotic and image processing.
K. S. Sim is a Professor with Multimedia University, Bukit Beruang, 75450
Melaka, Malaysia. He has won many International and national awards namely the Academic Science Malaysia (ASM) as Top Research Scientists
Malaysia (TRSM); He has filed more than 14 patents and more than 70
copyrights. He was also a recipient of the Japan Society for the Promotion of Science (JSPS) Fellowship from Japan in 2018. His research areas are signal
and image processing, noise quantization, image color coding, biomedical
imaging and biomedical engineering.
S. C. Tan received his PhD degree from Multimedia University, Malaysia.
He is now an Associate Professor at Multimedia University. His research areas are signal and computational intelligence, data classification, condition
monitoring, fault detection and diagnosis, pattern recognition and biomedical
disease data analysis.
Engineering Letters, 28:4, EL_28_4_46
Volume 28, Issue 4: December 2020
______________________________________________________________________________________
top related